Method and system for building digital twin models

The invention relates to a digital twin technology, and more particularly to a method and a system for building digital twin models.

Digital twin technology has gradually been applied to industry in recent years. Digital twin technology can be used to build a virtual model of a physical object, and there is connectivity between the physical object and the virtual model. The real-time data returned by a sensing unit can be serially processed, analyzed, and judged, so that the virtual model can generate feedback.

However, the amount of data of the virtual model is generally very large, and a huge amount of data computation is required to obtain the feedback result of the virtual model. Therefore, not only a huge operational processing resource is required, but also such a virtual model is not conducive to the evaluation of whether a component used as the physical object can be applied to machines of different specifications.

Therefore, an object of the present invention is to provide a method and a system for building digital twin models, and the method and system are capable of greatly reducing an amount of data of a virtual model corresponding to a physical object and speeding up the building of the virtual model.

Another object of the present invention is to provide a method and a system for building digital twin models, and the method and system are capable of greatly reducing a demand for operational processing resources.

Yet another object of the invention is to provide a method and a system for building digital twin models, a virtual model generated using the method and the system is conducive to the evaluation of whether a component used as the physical object can be applied to machines of different specifications.

A method for building digital twin models provided by one embodiment of the present invention is applicable for building a digital twin model of at least one component of a feeding system, the component has a set of actual eigenvalues and a set of actual eigenvectors corresponding to the set of actual eigenvalues, and the method for building the digital twin model is executed by at least one processor and includes following steps of: (A) receiving a user setting from a user interface, setting a geometric solid image according to the user setting, a contour of a simplified geometric solid of the geometric solid image corresponding to a contour of the component, and the user setting associating with a shape and a size of the simplified geometric solid; (B) sampling the simplified geometric solid of the geometric solid image to obtain second position data; (C) obtaining material data of the component from a database; (D) calculating a set of model eigenvalues and a set of model eigenvectors by a modal analysis method according to second size data of the simplified geometric solid, the second position data and the material data; (E) determining a similarity between the set of actual eigenvectors and the set of model eigenvectors by a modal verification method; and (F) when determining that the set of actual eigenvectors being similar to the set of model eigenvectors, defining the simplified geometric solid as the digital twin model of the component, and defining the set of model eigenvalues and the set of model eigenvectors as twin dynamic characteristics of the component. A data amount of the second size data of the geometric solid image is less than a data amount of first size data of the component, a data amount of the second position data is less than a data amount of first position data of the component, and the first size data and the first position data are stored in the database and are obtained from a component image of the component.

In some embodiments, the set of actual eigenvalues and the set of actual eigenvectors are calculated by the modal analysis method based on the first size data, the material data, and the first position data.

In some embodiments, the step (B) and obtaining the first position data from the component image are performed by a finite element method (FEM) or continuum mechanics.

In some embodiments, the step (B) includes following steps of: (B1) discretizing the simplified geometric solid into a plurality of second image blocks; and (B2) defining pixel coordinates of vertexes of the plurality of second image blocks as the second position data.

In some embodiments, obtaining the first position data from the component image includes following steps of: (G) discretizing the component image into a plurality of first image blocks; and (H) defining pixel coordinates of vertexes of the plurality of first image blocks as the first position data.

In some embodiments, the modal verification method is modal assurance criterion (MAC), mean phase deviation (MPD) or modal phase collinearity (MPC).

In some embodiments, the simplified geometric solid is a cube, a cuboid, a flat plate, or a cylinder.

In some embodiments, the component is a transmission element or a work platform, and the transmission element is a bearing, a ball screw, a rotary table or a linear guideway.

A system for building digital twin models provided by one embodiment of the present invention, includes at least one processor configured to execute the method for building the digital twin model.

In the following detailed description, many specific details are set forth in order to provide a thorough understanding of the invention. However, those of ordinary skill in the art will understand that the invention can be practiced without these specific details. In other cases, well-known methods, procedures and/or elements have not been described in detail so as not to obscure the invention.

Please refer to, illustrating a method for building digital twin models (hereinafter abbreviated as building method) and a systemfor building digital twin models (hereinafter abbreviated as system) provided by one embodiment of the present invention. The building method is executed by the system; and the systemis applicable for building a virtual model of at least one componentof a feeding systemof a machine through the digital twin technology, and finding out twin dynamic characteristics of the virtual model. The componentcan be, for example, but not limited to, a transmission element, such as a bearing, a ball screw, a rotary table or a linear slide, or a work platform. In order to clearly illustrate the spirit of the present invention, the following will take the work platformas an example of the componentfor illustration.

The systemcan be implemented in one server, or the systemcan be implemented in a distributed manner in multiple servers capable of communicating with one another. The systemincludes at least one processor and at least one storage device capable of communicating with each of the at least one processor. The systemis installed with several software applications, so that the at least one storage device, the at least one processor and the software applications in operation can be planned to cooperatively form a position sampling unit, a database, an unsimplified modal analysis unit, a geometric solid setting unit, a position sampling unit, a simplified modal analysis unit, a similarity determination unit, and a database. The position sampling unitand the unsimplified modal analysis unitare capable of communicating with the database; the geometric solid setting unit, the position sampling unit, the simplified modal analysis unit, the similarity determination unit, and the databaseare capable of communicating with one another; the similarity determination unitis capable of communicating with the unsimplified modal analysis unit; and the simplified modal analysis unitis capable of communicating with the database.

The method for building a digital twin model of the work platform(i.e., the building method provided by the invention) includes following steps, for example, but is not limited to the steps.

Firstly, in step S, the unsimplified modal analysis unitobtains first size data, material data, and first position data of the work platformfrom the database. The first size data, the material data, and the first position data of the work platformare pre-stored in the database, and the databasealso records corresponding relationships between the first size data, the material data, and the first position data. The first size data can be created or set by drawing a component image IMof the work platformby, for example, but not limited to, a drawing software application (for example, but not limited to, AutoCAD) installed in the system. The component image IMis a three-dimensional image, and an image Vof the work platformis displayed in the component image IM. The first size data can include, for example, but not limited to, data about a length L(for example, 730 mm) of the image Vin an axial direction D, data about a width W(for example, 375 mm) of the image Vin an axial direction D, data about a height H(for example, 170 mm) of the image Vin an axial direction D, and data about a perforation diameter and a groove depth of the image V, and each of the sizes is not limited to actual size or image scale size. The axial directions D˜Dare perpendicular to one another. The material data can include, for example, but not limited to, density and Young's modulus. The first position data can be obtained by, for example, but not limited to, sampling pixel coordinates of the component image IM.

The method for obtaining the first position data can be realized by a finite element method (FEM) or continuum mechanics. In the case of using the finite element method as an example as shown in, firstly, in step S, the position sampling unitobtains the component image IMof the work platformfrom the at least one storage device. Then, in step S, the position sampling unitspatially discretizes (i.e., dividing by grids) the image Vof the work platformin the component image IMto obtain a plurality of first image blocks B(or can be called sub-regions or elements) through a drawing software application (for example, but not limited to, AutoCAD) or a computer aided engineering (CAE) software application (for example, but not limited to, an analysis software application launched by ANSYS) installed in the system. A grid shape for the first image blocks Bis, for example, but not limited to, a triangle or a square. In this embodiment, the grid shape for the first image blocks Bis a triangle. Then, in step S, the position sampling unitdefines pixel coordinates of vertexes P(or called nodes or discrete points) of the first image blocks Bas the first position data of the work platform. Finally, the position sampling unitstores the first position data in the database.

After the unsimplified modal analysis unitobtains the first size data, the material data, and the first position data, the unsimplified modal analysis unitin step Scalculates a set of actual eigenvalues (i.e., actual eigenvalue data) and a set of actual eigenvectors (i.e., actual eigenvector data) of the work platformby a modal analysis method through a CAE software application installed in the systemaccording to the first size data, the material data, and the first position data. The set of actual eigenvalues is a plurality of natural frequencies of the work platform, and the set of actual eigenvectors is a plurality of modes of the work platform. The set of actual eigenvalues and the set of actual eigenvectors are dynamic characteristics of the work platform.

In step S, after a geometric shape, the first size data, the first position data, the material data (such as density and Young's modulus) and a density formula are known, an equation (1) related to the discretized image Vcan be obtained through the modal analysis method:[0}  (1)wherein [M] is a mass matrix, [K] is a rigid matrix, {u} is a set of displacements of the vertexes Pafter the discretization of the image V, and {ü} is a set of accelerations of the vertexes Pafter the discretization of the image V. Assume that the systemhas two vertices P, the mass matrix [M] and the rigid matrix [K] can be expressed as:

In order to solve the equation (1), {u} can be assumed to be {Ø}× sin(ωt+θ), and {ü} can be assumed to be

so the equation (1) can be simplified as:

wherein ωis a natural frequency of the image V, and {Ø}is a set of modes of the image V. Through solving a determinant det[K−ωM]=0, the natural frequency ωand a set of modes {Ø}can be found.

On the other hand, in step S, the geometric solid setting unitreceives a user setting from a user interface, and sets a geometric solid image IMcorresponding to a contour of the work platformaccording to the user setting, as shown in. The user setting associates with a shape type and a size of a simplified geometric solid Vdisplayed in the geometric solid image IM. The user interface can be provided by, for example, but not limited to, the geometric solid setting unitin conjunction with a CAE software application, and is displayed on a display device communicating with the at least one processor. For example, through an input device (for example, but not limited to, a keyboard, a mouse, or a touch panel of a display device) that communicates with the at least one processor, a user can select one (option of a cuboid) of shape options of a virtual model presented on the user interface according to a rough contour of the work platform(for example, the work platformlooking like a cuboid (i.e., shape type)); and according to the first size data (for example, but not limited to, the data about the length L, the width Wand the height Hof the image V) of the work platform, the user can also input dimensions (for example, but not limited to, a length Lof 730 mm in the axial direction D, a width Wof 375 mm in the axial direction D, and a height Hof 170 mm in the axial direction D) required by the simplified geometric solid Vwith a shape of a cuboid. The input of the shape and the dimensions is the user setting and is sent to the geometric solid setting unit. At this time, according to the user setting, the geometric solid setting unitis capable of defining (specifically, programmatically construct) the simplified geometric solid Vthat is cuboid-shaped and shown in a geometric solid image IM, as the digital twin model of the work platform, and defining the dimensions (that is, second size data) of the simplified geometric solid V. Since the simplified geometric solid Vis a simplified virtual model of the work platform, the shape and structure of the simplified geometric solid Vhave omitted many structural features (for example, but not limited to, perforation(s), groove(s) and rib(s)) in the work platformthat unlikely affect the dynamic characteristics. Therefore, a data amount of the second size data of the simplified geometric solid Vis far less than a data amount of the first size data of the work platform.

Then, in step S, the position sampling unitobtains the geometric solid image IMfrom the geometric solid setting unit, and samples (or discretizes) the geometric solid image IMto obtain second position data of the simplified geometric solid V. The sampling method can be realized by, for example, but not limited to, the finite element method or the boundary element method. Taking the finite element method as an example for sampling, as shown in, the position sampling unitdiscretizes the simplified geometric solid Vin the geometric solid image IMinto a plurality of second image blocks Bin step S, and then in step S, defines pixel coordinates of vertexes Pof the second image blocks Bas the second position data. In this embodiment, a shape of the second image block Bis a cube. However, in other embodiments, the second image blocks Bcan also have the same shape as the first image blocks B, and even a size of the second image blocks Bcan be the same as or different from a size of the first image blocks B. Since the shape and the structure of the simplified geometric solid Vhave omitted many structural features in the work platformthat unlikely affect the dynamic characteristics, a data amount of the second position data of the simplified geometric solid Vis also far less than a data amount of the first position data of the work platform.

Then, the simplified modal analysis unitobtains the material data from the databasein step S, and obtains the second size data from the geometric solid setting unitand the second position data from the position sampling unitin step S, and then adopts the same method as what the unsimplified modal analysis unituses, i.e., the modal analysis method, to calculate a set of model eigenvalues and a set of model eigenvectors of the simplified geometric solid Vbased on the second size data, the second position data and the material data. The model eigenvalues are natural frequencies of the simplified geometric solid V, and the model eigenvectors are modes of the simplified geometric solid V.

After the unsimplified modal analysis unitcalculates and obtains the set of actual eigenvalues and the set of actual eigenvectors, and the simplified modal analysis unitcalculates and obtains the set of model eigenvalues and the set of model eigenvectors, the similarity determination unitin step Sobtains the set of actual eigenvectors from the unsimplified modal analysis unitand the set of model eigenvectors from the simplified modal analysis unit, and then determines a similarity between the set of actual eigenvectors and the set of model eigenvectors by a modal verification method. The modal verification method can be, for example, but not limited to, a modal assurance criterion, a mean phase deviation method, or a modal phase collinearity method.

Taking the modal assurance criterion as an example for calculating the similarity, the similarity can be calculated by the following formula (5), wherein MAC(r,q) represents the similarity; Ørepresents a matrix of the set of actual eigenvectors; Ørepresents a matrix of the set of model eigenvectors;

represents a transposed matrix of the set of actual eigenvectors; and

represents a transposed matrix of the set of model eigenvectors. When the similarity is greater than or equal to a threshold value (for example, but not limited to, 0.8), it means that the set of actual eigenvectors is similar to the set of model eigenvectors. Conversely, when the similarity is less than the threshold value, it means that the set of actual eigenvectors is not similar to the set of model eigenvectors.

Then, in step S, when the similarity determination unitdetermines that the set of actual eigenvectors is similar to the set of model eigenvectors, it means that the current simplified geometric solid Vcan be equivalent to the work platform. At this time, the similarity determination unitin step Sdefines the simplified geometric solid Vas the digital twin model of the work platformand defines the set of model eigenvalues and the set of model eigenvectors as the twin dynamic characteristics of the work platform. In addition, in step S, the similarity determination unitalso tells the geometric solid setting unitto store the geometric solid image IMand its second size data in the database, tells the position sampling unitto store the second position data in the database, and tells the simplified modal analysis unitto store the twin dynamic characteristics and the material data in the database. And, the databasealso records corresponding relationships between the geometric solid image IM, the second size data, the second position data, the material data, and the twin dynamic characteristics.

Conversely, in step S, when the similarity determination unitdetermines that the set of actual eigenvectors is not similar to the set of model eigenvectors, it means that the current simplified geometric solid Vcannot be equivalent to the work platform, so that the similarity determination unitdo not define the simplified geometric solid Vas the digital twin model of the work platform, and do not define the set of model eigenvalues and the set of model eigenvectors as the twin dynamic characteristics of the work platform.

Through the procedure of the above steps Sto S, the present invention is capable of greatly reducing an amount of data of the virtual model and speeding up the building of the virtual model while greatly reducing a demand for operational processing resources, so that the virtual model is conductive to the evaluation of whether the work platformcan be applied to machines of different specifications.

Through the verification process of the above steps Sto S, the present invention enables that the virtual model with a less data volume can still be equivalent to the work platform.

Although the execution sequence of steps Sto Sin the above embodiment is independent of steps Sto S, the present invention is not limited to this example of the process. In other embodiments, steps Sto Scan be performed at any time point before step S.

In addition, although the above-mentioned embodiments use the work platformas an example for illustration, the systemand the method provided by the present invention can also be applied to building digital twin models and their twin dynamic characteristics of other components (for example, but not limited to, a screw) of the feeding system, or can also be applied to building digital twin models and their twin dynamic characteristics of components of other devices in the machine other than the feeding system.

Although the databasesandof the above embodiments are set up separately, the present invention is not limited to this implementation mode. In other embodiments, the databasesandcan also be integrated into one instead.

Although the specific embodiments of the invention are disclosed in the above implementation modes, they are not intended to limit the invention. The specification relating to the above embodiments should be construed as exemplary rather than as limitative of the invention, with many variations and modifications being readily attainable by a person having ordinary skill in the art to which the invention pertains without departing from the principles and spirit thereof as defined by the appended claims and their legal equivalents.